Yanhui Xie, Gayathri Seenumani, Jing Sun, Fellow, IEEE, Yifei Liu, and Zhen Li
Abstract This paper presents the development of a PC-cluster based real-time simulator for All Electric Ship (AES) Integrated Power System (IPS) analysis and optimization. The system is established at the University of Michigan with the nancial support from the U.S. Ofce of Naval Research (ONR). It is aimed to address the multi-disciplinary issues associated with the allelectric ship, such as optimal power management and dynamic system reconguration. In parallel with hardware construction, a modularized IPS model which includes the power generation module, the recongurable zonal electrical distribution system module, the ship propulsion module and the ship dynamic module is developed and integrated. Simulation GUIs have also been developed to provide a user friendly engineering environment where data acquisition and parameter tuning can be performed effectively. Both the system development efforts and preliminary simulation results for different shipboard operating scenarios are reported. Index Terms All Electric ship, Integrated Power System, Zonal Electrical Distribution System, Real-Time Simulator, Hardware-in-the-Loop.
enhance the reliability and survivability of the power distribution network, zonal electric distribution system (ZEDS) was introduced [5]-[6]. Unlike the convectional radial electric distribution system which radially distributes power to the loads through load centers, ZEDS employs two main buses (starboard bus and port bus) to provide redundant power ow paths for vital loads. With the introduction of Power Electronic Block Building (PEBB) [7]-[8], the ZEDS could seamlessly and dynamically recongure the power ow pathes in response to different priorities of loads for different real time battle scenarios.
I. I NTRODUCTION
EXT generation warships will be featured with integrated power system (IPS) which incorporates a set of primary and auxiliary power sources to provide the prolusion power and at the same time to energize the shipboard electric loads [1]-[2]. Coupled with the integration of power generation systems, the introduction of new technologies such as advanced power electronic systems [3], intelligent control systems and the state-of-the-art energy storage technologies [4] necessitates the real-time power management and power system reconguration, especially in situations of equipment failure and battle damage. The main objective of the power management is to ensure the continues power supply for the electric loads, thereby augmenting the reliability and survivability of the Shipboard Power System (SPS), and consequently enhancing the mission effectiveness of the warship. The power network associated with the All Electric Ships (AES) typically has electric propulsion, sophisticated electric weaponry systems and ship service as electric loads. To
This work was sponsored by the U.S. Ofce of Naval Research (ONR) under Grants No. N00014-03-1-0983 and N00014-05-1-0533. Yanhui Xie, Jing Sun, Yifei Liu and Zhen Li are with the Department of Naval Architecture and Marine Engineering, University of Michigan, Ann Arbor, Michigan 48109 USA (e-mail: yhxie@umich.edu; jingsun@umich.edu; yifeil@umich.edu; lizhen@umich.edu). Gayathri Seenumani is with the Department of Mechanical Engineering, University of Michigan, Ann Arbor, Michigan 48109 USA (email: gseenuma@umich.edu).
Fig. 1.
Fig.1 illustrates a one-line diagram of the IPS with the zonal architecture. The IPS is mainly comprised of power generation module (PGM), power conversion module (PCM), energy storage module (ESM), electric propulsion module (EPM) and vital/nonvital loads. Power generation module could be gas turbine, diesel engine or fuel cell. Reconguration of the power system shown in Fig.1 is achieved through the power electronic converters, which redirect the power ow and control the power distribution to individual loads. Since these power sources have different response time and the loads have different operating characteristics, dynamic optimal power management is critical to the power system
396
stability, efciency and performance. The power management system has to face several challenges brought up by the IPS, including: (1) How to coordinate the power converters with power sources to ensure uninterrupted power supply to vital loads? (2) How to recongure the system to achieve optimal efciency and maximum reliability? and (3) How to fulll the reconguration to ensure the power quality and system operation integrity during transients? To address the multi-disciplinary issues associated with the AES, advanced analytical and numerical tools are critical to enable the model-based investigation and system optimization. With the support from ONR, a real-time simulation platform for shipboard power system has been established at the University of Michigan to support the research and education activities related to AES initiatives. In the following sections, we will discuss the platform construction, IPS modeling and integration as well as GUI development. II. S IMULATOR HARDWARE CONFIGURATION
Fig. 2.
As described in the introduction section, the IPS is a large scale power electronics based electrical network system which contains many high frequency power switches. These switches are highly nonlinear and time consuming to simulate. Building a simulator for such an electrical system is always a challenge, given the system cost, required simulation accuracy and desired exibility. Other than analogy simulator which achieves real-time simulation by using scaled down analogue models of actual components, digital simulator becomes the trend due to its low maintenance cost and exibility. Many ofine simulation packages including MATLAB/Simulink , PLECS , SABER , etc., can perform ofine simulation, but they either cant interact with external hardware or their simulation speed are very slow for large scale system simulation. Real-time digital simulator is a promising approach since they avoid those drawbacks. In comparison with DSP and FPGA based real-time simulator, the PC cluster based simulation system would be better choice considering the low hardware cost, high simulation performance as well as the exibility provided by the modular system architecture. RT-LAB is a PC-cluster based expendable real-time simulator which is compatible with Matlab/Simulink , thereby allowing effective leverage of
commercially available MATLAB/Simulink toolsets, such as Control System Design and Analysis toolboxes, Code Generation toolboxes, and Physical Modeling toolboxes. Specialized tools such as ARTEMIS and RT-Events support multi-rate xed-time-step real-time simulation of power systems with dramatically improved computation speed and accuracy [9]. Fig.2 shows the conguration of the system. This system has 8 CPUs allocated in 4 physically separated targets. The CPUs in the same target exchange information through the shared memory while the different targets communicate through inniband switch with 10Gb/s speed. There are three host PCs which can talk with each target via 1Gb/s Ethernet switch. The targets can interact with the external hardware through 32bits PCI Bus I/O interfaces. Combining the FPGA event detection with specialized real-time interpolation algorithms toolbox RT-Events , the effective preciseness of the I/O could be better than 1s. The I/O interface provides a platform for data acquisition and signal conditioning modules that enable the implementation of high frequency analog/digital I/O, event capture, event generation. All of the targets and CPUs are synchronized either by software or by hardware. Thus all of the CPUs can synchronously interact with the analog and digital I/O. This feature make it possible for the system to implement physical components for hardware-in-the-loop simulation or to perform fast control prototyping experiments. Besides, the synchronized targets can run real-time simulation at different time step, making it very exible to distribute the complex model to different targets or CPUs within the simulation platform. This real-time simulator also can interact with other stand along RT-Lab targets through Ethernet. The stand along target could be used as experiment data acquisition device or any real-time simulation subsystems for AES. In parallel with hardware construction, the modularized model of IPS is developed and will be introduced in the following section. III. M ODULARIZED MODELING OF IPS Since the IPS is a large scale power system containing many high frequency power switches or other components whose simulation is very resource demanding, it is time consuming to simulate and debug such a large system ofine. On the other hand, parameters tuning of this complex system as a whole is a daunting, if not impossible, task. Furthermore, many of the subsystem in ZEDS are similar and can be reused. Therefore, modularized modeling approach is adopted for the IPS simulation. We split the whole IPS into power generation module (Gas Turbine and Fuel Cell), electric propulsion module and the ZEDS. ZEDS, by itself, consists of many power conversion modules and electric loads. A. Gas Turbine Module While many different types of power systems are used for shipboard applications, gas turbine/generator sets are quite often used as the shipboard prime mover. A combination of rst principles and empirical relationships have been used for the gas turbine model development [10]. The schematic of the
397
gas turbine model developed for the ship-board power system is shown in Fig. 3. The representation for the gas turbine model
reforming (ATR) for diesel fuel can also be incorporated. The fuel processing is typically slower as compared to the fuel cell. The FC-FPS model developed by J.Pukrushpan et.al. [11] was adopted, whose schematic is shown in Figure 5. The represen-
Fig. 5.
Fig. 3.
consists of the compressor, turbine, combustor models and rotational dynamics capturing the coupling between turbine and compressor. The compressor and turbine are modeled using empirical relations by regressing the corresponding performance maps [12]. The important dynamics modeled are the temperature and mass of the fuel burnt in the combustor and rotational speed of the turbocharger [13]. The gas turbine drives the power generation device which converts mechanical power to electrical power and generates three-phase AC power with 4160V L-L voltage and 60Hz frequency. For the particular system modeled in this platform, the total power output is between 255 KW to 1.4 MW. The open loop simulations
900 850 800 750 Open Loop Simulation of Gas Turbine Model
tation of the FC-FPS model consists of hydro-desulphurizer (HDS), catalytic partial oxidizer (CPOX), water gas shift (WGS) and preferential oxidizer (PROX), anode and stack voltage sub models. The HDS, WROX (WGS + PROX) are modeled as rst order lags. The important dynamics modeled are CPOX temperature, anode pressure and partial pressure of hydrogen in anode. The stack voltage model in [11] computes the voltage output of the stack as a function of the load current, anode and cathode pressures and membrane hydration. The output of the FC system is connected to the DC-Link in PCM4 through a DC/DC converter. For the particular system modeled, the total power output of the FC-FPS is between 80 KW to 330 KW. The open loop simulations for step changes
180 160 140 120 100 0 5 10 15 time sec 20 25 Open Loop Simulation of Fuel Cell Model
220
Power KW
700 650 600 550 500 450 0 5 10 15 20 time sec 25 Pdem PGT/gen 30 35
Fig. 6. Open Loop Simulations of Fuel Cell: Current Demand and Generated Power.
Fig. 4. Power.
for step changes in power demand are shown in Fig. 4 and the gas turbine has a fast second order response. B. Fuel Cell Module The Fuel Cell (FC), which is clean and highly efcient, is considered as an auxiliary power source for the IPS. The shipboard application uses fuel processing system (FPS) to get pure hydrogen from natural gas. Other models representing different fuel reforming technology, such as the auto thermal
in fuel cell load current are shown in Figure 6. It can be seen that the fuel cell has a slower response, compared to that of the gas turbine/generator set. The hydrogen starvation in the fuel cell due the the slow reformer dynamics is reected in the instantaneous increase followed by a dip in the power output. C. ZEDS Module Fig.1 shows the diagram of a two-zone IPS which has PGM, EPM, ESM and ZEDS. Since DC zonal architecture has many advantages over AC zonal architecture [6] for shipboard application, DC ZEDS was adopted in our model. The key
398
components of DC ZEDS are vital/non vital loads and power conversion modules whose diagram are shown in Fig.7 and Fig.8. The models for individual components are discussed briey in the sequel.
Vp4+ 5 1 S1
c 1+ 12-
2
2+
Vb2+
c 1+ 122+
S2 DC bus
3 DC bus Vb24
Vp4From PCM4 6
Switch2 Io L1 3 S3
c Io
1 Vo+
C1
R1
1+ 2+
1 To subbus
1- 2-
Vo2
Pulses
Vref Vf dbk
Vo
V1
Switch3 2 Vo
Controller
Fig. 7.
1) PCM1: Fig.7 shows the model of PCM1. PCM1 is a step down DC/DC converter with three recongurable switchboards. The step down DC/DC converter is modeled with the 1-leg Time-Stamped Bridge of the ARTEMIS toolbox while other components are modeled with SimPowerSystems toolbox. Manipulating the three switchboards can recongure power ow path of each electric zone. DC bus failure and recovery emulation also can be achieved by the manipulation of switchboards. The output voltage of PCM1 is 900VDC which is 200V less than the main bus. The loads of PCM1 are one nonvital load and one vital load under normal situation. One vital load will be added if the opposite main bus or PCM4/PCM1 is down because of either equipment failure or battle damage.
Upper input ABT
dropped only when its voltage decreased to 100V lower than the lower input. However, to balance load for the two DC bus, the upper input will take over again if its voltage is recovered to 50V higher than the lower input. 3) PCM3/6: Fig.8b shows the diagram of PCM3/6. PCM3/6 is DC/DC converter which is modeled with the 1-leg Time-Stamped Bridge of the ARTEMIS toolbox too. There is no ABT in PCM3/6 given the nonvital nature of the loads connected to it. The nonvital load will directly lose its power if the main bus or sub-bus on its side is down. 4) PCM4: Fig.8c is the diagram of PCM4 which conventionally is an AC/DC converter converting three-phase AC power to DC power by controlling the rectier ring angle. For our model, PCM4 has hybrid power sources, AC main bus and fuel cell, the output of AC/DC was connected with output of the DC/DC converter of fuel cell model through DCLink. The proportion of power drawn from AC bus and fuel cell respectively can be dynamically managed by splitting the desired current to the two input converters. To get well regulated DC voltage on the port bus and starboard bus, there are starboard side and port side output DC/DC converters drawing power from DC-Link and regulating the voltage on the two DC buses to 1100VDC. The modeling of the two output DC/DC converter is similar to the DC/DC converter in PCM1, the buck converter topology is adopted and modeled with SimPowerSystems Compatible 1-leg Time-Stamped Bridge of the ARTEMIS toolbox. Both of the output converters are regulated by PI controller. 5) Load: Vital/nonvital load was modeled as constant power load. All of the loads can draw certain amount of power from DC bus according to the command from energy management module. More complex load like DC motor or AC motor also could be modeled and integrated in the future. D. Propulsion Module
+ v -
Lower input
DC/AC
Load
Input
DC/DC
Load
(a)
(b)
AC Bus
DC-Link
DC/DC
Port Bus
Fuel Cell
DC/DC (c)
STBD Bus
Fig. 8.
2) PCM2/5: Fig.8a shows the diagram of PCM2/5. PCM2/5 is DC/AC inverter which is modeled with SimPowerSystems Compatible 3-leg Time-Stamped Bridge of the ARTEMIS toolbox. Since they energize the vital load which shouldnt lose power by any chance, there is an auto bus transfer (ABT) circuit which can automatically select power input port between the upper input and lower input. Usually the upper input has higher priority than the lower one and will be
1) Electric Propulsion Model: Electric propulsion model is a three-phase AC/DC/AC variable speed transmission system with the low speed, high torque Permanent Magnet Synchronous Motor (PMSM) driving the propeller. The AC/DC rectier is modeled with SimPowerSystems toolbox Universal Bridge. There is also a braking chopper on the DCLink to absorb the regenerated energy by the motor at the crash stop situation. The DC/AC inverter which works as the frequency converter and drives the propulsion PMSM is modeled with Time-Stamped Bridge of the ARTEMIS toolbox and controlled by a close loop speed controller. Except for threephase AC/DC/AC propulsion system, other AC propulsion technologies such as cycloconvertor [14], matrix converter [15] and high temperature superconductor (HTS) motor [16] also can be modeled and integrated into the propulsion modulein the future. 2) Ship Dynamic Model: The load torque to the eletric propulsion motor is determined by the ship dynamic model, which calculates the ship speed and propeller speed according to hydrodynamics. The ship model given in [17] is adapted. It includes the added mass and hydrodynamic forces and
399
moments acting on the ship. Given a desired ship speed, the desired motor speed and torque are calculated in this module and fed to the propulsion motor control unit. IV. M ODEL INTEGRATION , DISTRIBUTION AND
PRELIMINARY SIMULATION
stubline into signals ow on the arrowline. These bus model are incorporated to appropriate locations where it is necessary to split the model and assign different CPUs to the subsystems. Based on the integrated model, some preliminary simulation is performed which will be discussed in the following subsection. B. Preliminary Simulation
A. Model Integration and Distribution There are two stages for the IPS model integration. First of all, ZEDS module and propulsion module are integrated and tested respectively. As we discussed in the previous sections, the key components of ZEDS, loads and PCMs, are separately developed and tested. After that, all of the PCMs and loads are interconnected and form the two zones ZEDS. The integration of ship dynamic model and propulsion model is quite straightforward too. The desired propeller torque and speed signals which is calculated by ship dynamic model are sent to the motor in the propulsion model. Then ZEDS and propulsion modules are connected with the power generation module. The IPS is a large scale system which has many subsystems with different characteristics. For example, the dynamic of G/T and FC are relative slow, 1ms time step is sufcient. On the other hand, PCMs have high frequency power switches, the subsystem time step is 50s in our case which is much shorter than PGMs. To get a relative balanced computation task distribution among all CPUs for better simulation performance, one has to distribute the model properly into the 8-CPU simulator. There are several considerations that need to be taken into account in allocating resource: (1) The subsystem assigned to CPU should not be too complicated, otherwise overruns will result in poor simulation performance; (2) Since realtime communication between targets is a resource demanding process, it is advantageous to assign a modularized subsystem as one entity to one CPU instead of splitting the subsystem into several parts to reduce the number of signals to be exchanged between targets; (3) Each CPU can run real-time simulation with one sampling rate, subsystems with slow sampling speed can be grouped with the subsystem with fast sampling speed. Given these constraints, we assign one CPU for subsystems PGM(Gas Turbine), PGM(Fuel Cell) and EPM while splitting the ZEDS into two subsystems with two associated CPUs. Additional subsystems such as energy storage subsystem and power management subsystem will be developed and assigned to left CPUs in the future. Another challenge for model distribution is that all of the distributed subsystems exchange information through RTLAB block-OpComm which is only compatible with arrowlines. Arrowline indicates the unidirectional signal ow and is the default connection for Simulink as well as many Simulink compatible toolboxes blocks. However, the power conversion and propulsion subsystems developed with SimPowerSystems toolbox use stublines by default. The stubline passes signals bidirectionally just as a physical wire would do. The incompatibility between SimPowerSystems sub models and OpComm blocks is resolved by customer built DC and AC bus models which convert the signals ow on the
15
4
20
25
30
35
40
45
15
4
20
25
30
35
40
45
15
4
20
25
30
35
40
45
Scenario1
15 20 25
Scenario2
30 time, [s]
Scenario3
35
Scenario4
40 45
Fig. 9.
Fig.9 shows power consumed by loads during a failure and reconguration process where four scenarios are represented. Scenario 1: the port bus or local PCM1/PCM4 is down, the non vital load lose its power while the vital load draw power from the STBD bus. Scenario 2: the port bus is recongured or the PCM1/PCM4 is recovered, all of loads draw equally 40KW power from both bus. Scenario 3: the STBD bus is down which leads to the STBD side nonvital load loss its power. The vital load works ne since it is switched to draw power from the port bus. Scenario 4: both buses are down, consequently all of the loads loss their power. This testing veried the failure emulation and reconguration capability of the ZEDS model, which is critical to enable the algorithm development of intelligent reconguration of ZEDS in the future. Fig.10 shows the transient of propulsion motor when the ship speed is accelerated from 0 knot to 8 knots. To boost the propeller speed, the actual torque is signicantly larger than the desired torque during the starting up period. Both the actual torque and speed curves match with the desired curves very well after that, which conrms that the propulsion model captures the ship dynamic model as required. V. G RAPHICAL USER INTERFACE (GUI) DEVELOPMENT Since the system is very complicated and many signals have to be sent to the console on the host PC, a single window console is too busy to be used to monitor the signals during the simulation. On the other hand, we have to interact with the real-time simulated model to operate the recongurable
400
x 10 15 Torque, N.m
10
Numeric boxes next to the PCMs allows manual inputs to change the loads associated with corresponding PCMs. The ship speed command knob controls ship speed. The two waveform charts display the desired and actual speed and torque for the propeller.
300
50
100
150
200
250
50
100
200
250
300
Fig. 10.
switches in the PCM1. So GUI is necessary to create an engineering environment where one can monitor and manipulate the real-time simulated model. We developed the GUI based on Testdrive V2.1.3 which is compatible with both MATLAB/Simulink and LabVIEW . The GUI has a multi-level structure. The top level GUI is similar to the structure shown in Fig.1, displayed signals can be used to indicate the health and status of the system. We are also able to navigate to each of the sub-level GUIs for different subsystems by clicking on their respective button. In the sub-level GUI, data acquisition and control signals of the subsystem are displayed.
Failure, Reconfiguration, Operation and Drive Scenarios
ZEDS: Failure, Reconfiguration and Operation Scenarios
PCM 1
PCM 1 400 300 200 PCM 3
VI. C ONCLUSION The preliminary simulation results verify that the AES/IPS simulator can perform simulations for scenarios such as failure emulation, power ow path reconguration and energy management. The simulator incorporating different I/O hardware can support other system development activities such as hardware-in-the-loop simulation and fast control prototyping. This simulator provides an effective platform for AES system research. At this moment, there are two ongoing projects using this simulator, one involves optimal reconguration energy management, and another deals with hybrid power system coordination control. A hybrid power system testbed with fast control prototyping capability is also under construction. R EFERENCES
[1] N. Doerry and J.C. Davis, Integrated Power System for Marine Applications, Naval Engineers Journal, 1994. [2] N. Doerry, H. Robey, J. Amy and C. Petry, Power the Future with the Integrated Power System, Naval Engineers Journal, 1996. [3] T. Ericsen, N. Hingorani, and Y. Khersonsky, Power electronics and future marine electrical systems, IEEE Transactions on Industry Applications, 2006. [4] C.A. Luongo, T. Baldwin, P. Ribeiroand, and C.M. Weber, A 100 MJ SMES demonstration at FSU-CAPS, IEEE Transactions on Applied Superconductivity, 2003. [5] C. Petry and J. Rumburg, Zonal Electrical Distribution Systems: An Affordable Architecture for the future, Naval Engineers Journal, 1993. [6] J.G. Ciezki and R.W. Ashton, Selection and stability issues associated with a navy shipboard DC zonal electric distribution system, IEEE Transactions on Power Delivery, 2000. [7] T. Ericsen and A. Tucker, Power Electronics Building Blocks and potential power modulator applications, IEEE Conference Record of the Twenty-Third International Power Modulator Symposium, New York, 1998. [8] Z. Ye, K. Xing, S. Mazumder, D. Borojevic and F.C. Lee, Modeling and control of parallel three-phase PWM boost rectiers in PEBB-based DC distributed power systems, IEEE Applied Power Electronics Conference, 1998. [9] C. Dufour and J. Belanger, A PC-based real-time parallel simulator of electric systems and drives, International Conference on Parallel Computing in Electrical Engineering, 2004. [10] G.G. Kulikov and H.A. Thompson, Dyanmic Modeling of Gas Turbines, Springer, 2004 [11] J.T. Pukrushpan, A.G. Stefanopoulou and H. Peng, Control of Fuel Cell Power Systems, Principles, Modeling, Analysis and Feedback Design, Springer, 2004. [12] P. Moraal and I. Kolmanovsky, Turbocharger modeling for automotive control applications, Journal of Engg. For Gas Turbines and Power, 2003. [13] V. Tsourapas, J. Sun and A. Nickens, Control oriented modeling and analysis of a hybrid solid oxide fuel cell and gas turbine (SOFC/GT) system, Proceedings of the 2006 Fuel Cell Seminar, 2006. [14] E.A. Lewis, Cyclo convertor drive systems, Sixth International Conference on Power Electronics and Variable Speed Drives, 1996. [15] P. Tenti, L. Malesani and L. Rossetto, Optimum control of N-input K-output matrix converters, IEEE Transactions on Power Electronics, 1992. [16] S.S. Kalsi, B.B. Gamble, G. Snitchler and S.O. Ige, The status of HTS ship propulsion motor developments, IEEE Power Engineering Society General Meeting, 2006 [17] T.I. Fossen, Marine Control Systems: Guidance, Navigation and Control of Ships, Rigs and Underwater Vehicles, Marine Cybernetics AS, 2002.
Drive Scenarios
Torque Act Cmd
40
KW
PCM 3
40
KW
100 0
0
PCM 2
20
40
Time
60
80
100
40
KW
PCM 2
40
KW
Speed 6
Act Cmd
PCM 4
PCM 4
PCM 5
40
KW
PCM 5
40
KW
0
20
40
60 Time
80
100
PCM 6 4 0
KW
PCM 6 4 0
KW
PCM 1
PCM 1
Fig. 11.
As an example, the Failure, Reconguration, Operation and Drive Scenarios, a sub-level GUI for power management, is given in Fig.11. The Failure, Reconguration, Operation and Drive Scenarios GUI is the one which can set the ship speed, emulate PCM or DC bus failure and recongure power ow path. The buttons on the GUI, corresponding to Switch1, Switch2 and Switch3 of PCM1 in Fig.7, can be pushed on or off to emulate bus crashing or to redirect the power ow.
401